Keyword research for SaaS, step by step
Updated
For a new SaaS with no backlinks, keyword research is not about finding the biggest word. It is about finding a small group of words that real buyers type and that a new site can rank for. Here is the process we use, with the real numbers from our own Idea Check page.
Step 1: Start from the problem, not your product name
Nobody searches for your product name yet. They search for the job they want done. Write down the problem in the words a buyer would use: 'validate a business idea', 'convert image to dxf', 'resume keyword checker'. Aim for 5 to 10 seed phrases.
If you are still choosing what to build, the Reddit digests show the problems people complain about in their own words, which makes good seeds.
Step 2: Expand with Google autocomplete
Type each seed into Google and write down every suggestion. Then add a letter, a preposition or a modifier after it: 'for', 'free', 'online', 'template', 'vs', 'how to'. This is free and shows phrases real people typed.
Keep going until you have around 50 phrases. Do not filter yet.
Step 3: Get volume and difficulty for every phrase
Paste the list into a keyword tool that gives monthly searches and a difficulty score (Semrush, Ahrefs, or a free tool like Google Keyword Planner for volume only). Record three columns: monthly searches, difficulty (KD), and intent.
- Small volume is normal. When we did this for our own Idea Check page, nothing that directly described the product passed 1,000 searches a month, and many phrases showed only 10 to 30.
- Some tools show no difficulty for low-volume phrases. Treat a blank as 'unknown', not as zero.
- Mark intent. A phrase like 'idea validation tool' means the person wants a tool. 'what is product market fit' means they want a definition. Prefer the first kind for a SaaS.
Step 4: Look at page 1 before you commit
Volume and difficulty are estimates. Page 1 is the evidence. Search each candidate and look at who ranks.
- Forum threads, Reddit posts and thin blog posts: a good sign. Nobody has written a proper answer.
- Small tools and indie sites: fine. You can win by being better or narrower.
- Only big brands, directories and Wikipedia: skip it and look for a narrower phrase.
Also check whether the results match what you would build. If page 1 is all definitions and you sell a tool, the intent does not fit and the traffic will not convert.
Step 5: Pick a short list and give each word its own page
Choose about 5 core words: the ones with a real buyer intent, low difficulty and a page 1 you could beat. Then plan one page per word. Two pages aiming at the same phrase compete with each other and neither ranks.
- Put the main word in the title and the H1 of its page.
- Write the answer people need, with real steps, not a sales page. Link to the product once, where it helps.
- Start with the lowest difficulty words. A first ranking page also gets the site known to Google.
Re-check after a few weeks in Google Search Console. The words you actually appear for are often different from the ones you aimed at, and those are the best clues for the next page.
Do it by hand or have it done
All five steps can be done by hand in an afternoon with Google and one keyword tool. IdeaPanner walks through the same checks for free. If you would rather have a person do it, the Idea Check is done by hand with real data, not an AI score. For the full checklist beyond keywords, see how to validate a SaaS idea.
Questions
How many keywords do I need for a new SaaS?
About 50 to collect, about 5 to commit to. Each core word gets its own page.
Should a new SaaS target high-volume keywords?
Usually not. A new site with no backlinks rarely ranks for big words. Start with low-difficulty, specific phrases and build up.
What if a tool shows no difficulty score?
That is common for low-volume phrases. Treat it as unknown and judge by what ranks on Google page 1.
Is free keyword research enough?
Autocomplete, Google Trends and Search Console are free and cover a lot. A paid tool adds volume and difficulty numbers, which save time but are not required to start.